Artificial intelligence as cognitive scaffolding in the engineering design process: A STEM education perspective

Abstract

Artificial intelligence (AI) is increasingly used in STEM education; however, limited evidence explains how it functions as cognitive scaffolding within the engineering design process. This study examined the effects of AI-supported scaffolding on students’ problem-solving competence, engineering design performance, and redesign competencies. A quasi-experimental pre-/post-test control group design was used, involving 124 lower-secondary students completing a six-week hydraulic robotic arm project. The experimental group received AI support during solution selection and prototype improvement, whereas the control group received conventional instructional resources. Data were collected using an eight-dimensional engineering design rubric and analyzed using t-tests, analysis of covariance, and effect size statistics. After controlling for baseline differences, the experimental group significantly outperformed the control group in problem-solving competence, F(1, 121) = 30.35, p < .001, partial η² = .348; engineering design performance, F(1, 121) = 42.43, p < .001, partial η² = .420; and redesign competence, F(1, 121) = 16.94, p < .001, partial η² = .118. The strongest improvements were observed in solution selection, problem understanding, scientific reasoning, and reflective reasoning. These outcome-level findings are consistent with the proposed role of conversational AI as adaptive cognitive scaffolding in STEM education. However, because the student-AI interaction processes were not directly measured, the results should not be interpreted as an empirical verification of the scaffolding mechanism.